Artificial intelligence has been the most exciting and contentious topic in technology for more than half a decade. A rising chorus glorifies its potential for change, while others point to the likely impact of AI-borne disruption on employment and career paths.
Whether you’re pro, con or somewhere in between, the convergence of technologies like machine learning and predictive analytics looks set to define the next phase of finance transformation. CFOs have been on a digital journey that began in the 90s with ERP and led to the present-day adoption of corporate performance management (CPM). Like its predecessors, I believe ‘AI’ is a new technological wave that will wash away old ideas about how finance adds value.
How do we define AI and machine learning? McKinsey says it’s “the ability of machines to exhibit human-like intelligence — for example, solving a problem without hand-coded software containing detailed instructions.” In finance, AI means emerging systems that can automate processes in an intelligent way, and predict future outcomes based on past data.
The drivers pushing AI forward are common across industries: eliminating manual processes, making business more nimble and responsive, and supplying insights in real-time for data-driven decision making. It’s the first and last on that list that makes AI so compelling for finance.
That’s because expectations for the CFO’s role – in fact the entire finance function – are on the rise. Boards expect greater detail and context in reporting and they want it faster and more frequently. They also want CFO’s to be a source of business insight rather than the company’s chief bean counter. AI and machine learning promise to deliver new tools that make it easier to meet those expectations.
From my experience, there will be four primary use cases for AI in finance:
1. Digital assistants
Smart assistants like Apple’s Siri can already handle questions on stock quotes and weather. What if you could ask a bot on you mobile for the latest financial results, and flag up any area that’s fallen short of, or exceeded, forecast? Effectively you’d be having a conversation with the CPM system, and could ask it to tell you where the anomalies occurred (a product, a region, both?), without having to open a dashboard or query a database.
Think this is far-fetched? Our parent company UNIT4 is already trialling its WANDA digital assistant which delivers just these kinds of capability.
2. Automated forecasts
While there are already solutions that bring a level of automation to forecasting, statistical methods vary widely and it can be a challenge to select the most appropriate method for the outcome you want to predict. This is an area where AI can help. In fact there are trials going on right now where machine learning is being applied to actuals for the previous six months to predict results for the 7th – with under 5% deviation.
3. Error-free data entry
Even when bespoke input templates have been created, data entry remains stubbornly prone to error. What if your CPM system could give you automatic feedback while you were typing, when it noted a conflict or an aberrant figure that simply seemed out of place when compared to current and historical results or normal calculations? While this capability is arguably achievable now, defining the rules would have to be done manually for each organisation, and would take ages.
4. Automated analytics
When a delta pops up in any summary or analysis of results, the deviation has to be investigated and its source determined in detail. Current finance systems allow you to create lists of current deviations but drilling down into the figures is still a manual process that takes time to address. In the near future, AI will be able to do this for you. Perhaps the analysis will happen through a digital assistant much like the previous example – ‘Wanda/Siri can you explain the source of the deviation on line three?’
As advanced technologies like these continue to penetrate the finance function, new data analysis skills must be part of the development and acquisition of finance talent. The more transactional roles in finance teams will become less common, as the need for strategic thinkers with cross-functional knowledge and technology mastery goes up.
AI may be knocking loudly on the door, but most of the solutions hyped today as ‘AI’ fall short of addressing CFO’s biggest challenges. A recent report from the Shared Services & Outsourcing Network suggests that most things labelled AI today are “limited cognitive solutions centred around a very narrowly defined knowledge domain.”
A lot of machine learning happens on so-called neural networks, which find solutions by trying hundreds of thousands of combinations to solve a problem, then present you with the best one. The system however can’t tell you why it’s the best one.
It’s the combination of machine learning and better predictive analytics within performance management that will likely deliver the breadth of practical benefits the office of finance needs from AI. The critical mass effect that kicks off widespread technological adoption won’t happen without it the convergence of both technologies.
How to prepare?
The CFO role is evolving and AI is going to accelerate the transformation. To prepare, I would recommend that anyone in a finance leadership role understand the key emerging technologies and start making decisions about the optimum time to run pilots or other initiatives to test new innovations. Determining the data-centric people skills and capabilities you’ll require must be part of this process.
Technology is changing so rapidly that you can’t blame CFOs for being cautious about AI. Maybe a wait-and-see attitude makes sense – we all need to better understand what shape the solutions for AI in corporate finance are going to take. The risk is in waiting too long, and potentially missing out on capabilities that are so advantageous, not adopting them puts you at risk of competitive disadvantage.
Matthias Thurner is CTO of prevero.